Partitioned Model vs Parallelized Solver Available to Purchase

Nowadays, engineers still search for more efficient methods in order to decrease simulation times. However, most simulation environments do not use the full power provided by modern PCs. Even though every modern computer is equipped with a multicore processor, only very few simulation environments use more than one core for simulations. There are various possibilities to parallelize simulations. One approach is to partition the model into several submodels. Using adequate solvers for each submodel can result in lower computation times, especially if there is a significant difference in the time constants of the submodels. Other approaches are based on parallelization of the ODE solver. For example, it is possible to parallelize the linear algebra methods inside the solver. Parallelization of the solver itself is another way to use the multicore architecture. From the modeling and simulation point of view, the latter approach is more interesting. Consequently, the question is whether it is beneficial to partition the model or to use a parallelized solver. In this paper this question is answered at least for an example system. However, the more efficient approach may not be the better approach for the usage inside a simulation environment. Therefore, it is discussed which approach can be automated and integrated easier into a simulation environment.